Projects

Set asset allocation and fixed/variable withdrawal rates, and visualize historical outcome paths for 30-year retirements 1928-1991. Further discussion in Advisor Perspectives.

Explore an MTA dashboard data using Plotly Dash, aggregating from turnstile-level data in DuckDB, demonstrating the performance of a column-oriented data warehouse. dbt data pipeline for ingestion and initial transformation. Further discussion in Numbers With Wings: The Modern Data Stack-In-A-Box.

Search for pizza (or coffee, or ice cream) in selected local NY areas, combining Google, Yelp, Foursquare data into a single Bayesian ranking.

Generate fund names using RNN word completion trained on a dataset of existing funds (blog post).

Visualizing the Fintwittersphere, organized by force graph and other neighborhood algorithms using follower and topic similarity

Basic portfolio optimization using historical asset class performance data.

Fast distributed hyperparameter tuning.

Primer on threshold selection for classification.

Intro to reinforcement learning with trading examples.

Using XGBoost classification to predict stock performance quintiles (a somewhat failed / overfitted experiment)

Using Keras and sklearn regression to predict stock performance (low R-squared but with right data sometimes gets results)